Measuring Availability in Service Level Agreements

Dear All,

Availability is one measure of quality used in Service Level Agreement programs by telecommunication providers today. The way availablity is measured by vendors determines the compensation customers receive for service outages and determines how good the guarantee is. The way the measures are done today is via time series, i.e. measure availability over time. This doesn't take into account the network size. The only way to take into account network size is to measure availability across the network. Doing so provides compensation based on the amount of the network that was not available. Please see below for a complete, but concise, summary of what the different availability measures actually guarantee! Thanks

Measures of Availability

Availability of service is measured today using time series data by all vendors in the telecommunications industry, for all their services. Cross section data is not used today to measure availability. But cross section data is very useful in quantifying a customer's network availability.

The difference between the times series data and cross section data is:

1. Times series measures the availability of an individual IP-VPN site, Frame/ATM port or pvc, or Private line circuit over time.

1. Cross section data measures the availability of an IP-VPN Network, Frame/ATM Network or Private Line Network at a point in time.

In other words:

1. Time series measures Site Availability, Port Availability, PVC Availability, or Circuit Availability for IP-VPN, Frame/ATM and Private Line service respectively. These measures are used in SLA programs today.

1. Cross section data measures Network Availability for a IP-VPN Network, Frame/ATM Network or Private Line Network. These measures are not used in SLA programs today. (The name "Network Availability" is used to describe availability measures by some vendors in their SLA programs. But the measurement is a time series measure, a measure of availability on a site, port or pvc over time, not a measure across the network.) The cross section approach is a patent pending methodology of L. Swanzer - E2E SLA Support, LLC. Their use requires license be obtained by the telecommunications providers to offer these metrics to their customers.

The difference between a time series measure of availability and a cross section measure of availability is shown in the table below (please see

if you can not read the table. This memo is on our website). The time series approach measures the availability of each IP-VPN Site (it could also be a Frame/ATM port or pvc, or a Private Line circuit) for any given month. So for example, sites 1, 2 and sites 6-10 had 100% availability in January. Sites 3, 4 and 5 had 99.03% availability in January. Sites 3, 4 and 5 had a 7 hour outage, from 2pm-9pm on January 15th, which implies 99.03% availability.

Jan Jan Jan Jan Sites 1/1-1/15 1/15 1/15-1/31 Site Avail. %Sites

Site 1 100% 100% 100% 100% 10% Site 2 100% 100% 100% 100% 10% Site 3 100% 0% 100% 99.03% 10% Site 4 100% 0% 100% 99.03% 10% Site 5 100% 0% 100% 99.03% 10% Site 6 100% 100% 100% 100% 10% Site 7 100% 100% 100% 100% 10% Site 8 100% 100% 100% 100% 10% Site 9 100% 100% 100% 100% 10% Site 10 100% 100% 100% 100% 10%

Total 100% 70% 100% 99.71% 100%

Percent Time in 48.39% 0.97% 50.64% 100% Range

The cross section approach measures availability of the network at various points in time. From January 1st to January 15th, there were no outages on the network. Therefore the network had 100% availability during those first 15 days of the month. On January 15th, from 2pm to

9pm, sites 3, 4 and 5 had 0% availability (these sites were affected by a seven hour outage). Therefore, from 2pm-9pm, on January 15th, the network availability was 70%.

Average availability is sometimes used as the SLA metric for availability. The average availability is the same whether the measure of availability is Network Availability or Site Availability. The method used to calculate average availability today is by averaging the individual Site Availabilities. The average availability measures are a weighted average. With Network Availability the weights are the time spent at different ranges of availability, with Site Availability (used today) the weights are the percent of Sites with different levels of availability. The Average Availability, in this case, was

99.71% shown in the bottom right corner.

Network Availability, Site Availability and Average Availability all have their own strengths and weaknesses.

The weakness with Network Availability is that compensation is capped by the amount of time the network is below the target level of availability. So, if 50% of the network is down for a couple of hours then the most the customer can be credited is the percent of time the network was down. For a 2 hour outage this amounts to .28% (2 hours out of 720 total hours in the month) of the total network charges. Even if the entire network is down, the credit is capped by the percent time the network was down. So if 100% of the network is down for 2 hours the maximum compensation due the customer is .28% of the network charges. On the other hand, Site Availability provides compensation based on the percent of the sites that are down. So if

50% or 100% of the Sites were down for 2 hours Site Availability would compensate a customer up to 50% or 100% of the network charges, respectively.

The weakness of Site Availability is that compensation is capped by the percent of sites that are below the target level of availability. For example, if there were a problem, in a given month, where a site had recurring problems that lasted much of the month, the customer's compensation would be capped to be a maximum of the charges on the problem site. For a large network, say a network with 100 sites, this would amount to a 1% of the total network charges are eligible for credit. If, on the other hand, Network Availability were the measure of availability, then the customer would be eligible for up to a maximum of 100% of the network charges, if the site caused the network to be below the target level for the entire month.

In general, Site Availability is going to compensate better then Network Availability when outages are of relatively short duration and impact many sites all at the same time. Network Availability will provide more compensation to the customer in months when outages are sporadic, occurring at different points and time during month.

Finally, Average Availability is an all or nothing deal. It provides either the best compensation or the worse compensation of the three metrics, depending on the outages. With Average Availability compensation is not limited by the amount of time the network is below the target level of availability nor is it limited by the percent of sites that are below a targeted level of availability. If the Average Availability target is missed the customer is eligible for compensation for all the network charges.

The issue with Average Availability is that it won't provide any compensation at all if the target is met, even though there could have been outages on the network that required compensation under Network Availability and/or Site Availability. If the target for Average Availability is 99.99% then a network wide outage that last 4 minutes and 31 seconds would not be eligible for any compensation. Similarly an outage that lasts 72 hours at a single site, on a network with 1000 sites, would not be eligible for any compensation. In both cases, the

4 minute 31second network wide outage and the 72 outage on a single site, would meet a 99.99% average availability target.

None of the availability measures protect the customer against all types of outages. For complete coverage against any type of outage you may consider combining Site Availability and Network Availability. For example, include both measures of availability in the Service Level Agreements. The measure that applies in any given month is the metric that provides the most compensation to the customer. Another way of getting the benefit of Site and Network Availability is to multiply the percent of the time of the month a particular outage lasted by the percent of total sites effected. The result would be the total percentage of the network that was affected by the outage: Percent time of the month * Percent of Circuits. Applying the resulting percentage to total network charges would amount to a refund of the customer's charges for the effected circuits, for the time the circuits were down. Of course an additional percent credit would have to be added to compensate the customer for the burden of the outage.

This is a very brief summary of today's measures of availability. We have a detailed presentation, which you can view from our website. You can also find out more about our: SLA administration services (tracking, reporting, claim processing), consulting services, presentations, and other services. Please see

for more information.

Network Availability and the variants of it are patent pending business methodologies of L. Swanzer and E2E SLA Support, LLC.

Thank you,

Larry Swanzer E2E SLA Support, LLC

908-806-4097